Abstract:
"Many cinnamon growers are unaware of diseases affecting their crops, leading to significant yield losses. Leaf chlorosis and rough bark disease are particularly problematic, often mistaken for natural aging or overlooked due to lack of knowledge. Additionally, the presence of leaf gall mites can impact cinnamon leaf oil yield and quality. Lack of preventive measures exacerbates the problem, as farmers struggle to identify and address these pests effectively. The absence of comprehensive image datasets further hinders research and diagnosis efforts. Establishing such datasets could greatly benefit the research community, facilitating automated inspections and ultimately improving cinnamon harvests. Increased awareness and access to resources for disease prevention are crucial for mitigating yield losses and ensuring the sustainability of cinnamon cultivation.
While numerous studies have focused on leaf disease detection, there's a notable absence of research specifically targeting cinnamon leaf diseases. This study aims to address this gap by proposing a deep ensemble neural network approach for leaf disease classification in Real time. Utilizing a dataset created by the author from plantation images. Author has managed to develop custom CNN for the created dataset with accuracy above 99%. Furthermore, the author experimented with the dataset by applying transfer learning with various state-of-the-art CNNs. Finally, the author applied a DL ensemble method by leveraging transfer learning models.
This real-time classification capability represents a significant advancement, providing growers with an accessible and practical solution for disease detection. Ultimately, this research lays a foundation for more robust, automated crop inspection, contributing to improved disease management and the sustainability of cinnamon cultivation."